A Comparative Study of Bird Species Classification Using K-Nearest Neighbors, Convolutional Neural Networks, and Support Vector Machines

Abstract

Bird species classification is crucial for ecological monitoring and biodiversity conservation, providing valuable insights into ecosystem health. This study aims to compare the performance of three machine learning algorithms—K-Nearest Neighbors (KNN), Convolutional Neural Networks (CNN), and Support Vector Machines (SVM)—in classifying bird species from images. A dataset of 15 bird species was selected, and data augmentation techniques were applied to enhance model generalization. KNN, CNN, and SVM were implemented using pre-trained MobileNet for feature extraction and evaluated based on accuracy, F1-score, and robustness to dataset variations. Results showed that CNN outperformed both KNN and SVM, achieving an accuracy of 92% and an F1-score of 0.92, while KNN achieved 71% accuracy and an F1-score of 0.71, and SVM performed poorly with only 24% accuracy and an F1-score of 0.24. The CNN model's superior performance highlights its effectiveness for complex classification tasks, making it the most suitable for bird species classification in ecological applications. Based on these findings, it is recommended that CNNs be used for large-scale biodiversity monitoring. Future research should explore advanced CNN architectures and combine them with techniques like transfer learning to improve performance further. Additionally, KNN may benefit from integrating more profound feature learning methods, while SVM may not be ideal for such tasks.